Objective: Flexion and erect standing radiographs were evaluated in the current study to compare their sensitivity in detecting articular cartilage wear.
Design: Prospective cohort study.
Setting: A tertiary care hospital outpatient orthopedic clinic.
Patients: All patients with osteoarthritis of the knee ages 40 to 75 scheduled for arthroscopic debridement between March 1995 and November 1997 were considered for the current study.
Intervention: Radiographs were obtained 1 week preoperatively in both the 3-foot standing anteroposterior (AP) and a 45° posteroanterior (PA) flexion weight-bearing projection. Joint space height was measured with a ruler in millimeters at the narrowest point of each compartment. All radiographs were assessed by two independent observers who were blinded to the arthroscopic findings and clinical symptoms of the subjects.
Main Outcome Measures: Prediction accuracy of each radiograph for severe Grade IV articular cartilage wear in tibiofemoral compartments.
Results: One hundred fifty-two patients with a mean (± SD) age of 60.5 ± 8.5 years were enrolled in the study. Fifty-one percent were female. Twelve patients were categorized as having severe lateral compartment articular chondropathy (Grade IV) at the time of arthroscopy. The lateral joint space height averaged 1 .0 ± 1.7 mm SD on the 45° PA radiograph compared with 2.7 ± 1.1 mm SD on the 3-foot standing AP view. Using a cutoff of 2 mm or less, the 45° PA view was much more sensitive (83% versus 42%) at correctly detecting the most severe chondropathy. Forty-one patients were classified with severe Grade IV medial compartment chondropathy at arthroscopy. There was little difference in the average joint space height measured by the 45° PA view (1.4 ± 1.4 mm SD) or the 3-foot standing AP view (1.9 ± 1.6 mm SD). A number of cutoff measures were evaluated, but no significant advantage could be found for either view in evaluating the medial compartment severity.
Conclusions: The bilateral 45° PA is superior for detecting lateral compartment wear but offers no advantage on the medial side. This view should be considered as the screening radiograph of choice in evaluating osteoarthritis of the knee.
Ottawa Hospital–General Site, and the University of Ottawa Departments of Surgery and Community Medicine, Ottawa, Ontario, Canada
Received November 15, 1999; accepted October 6, 2000.
Address correspondence and reprint requests to Geoff F. Dervin, MD, 5004-501 Smyth Rd., Ottawa Hospital-General Site, Ottawa, Ontario, K1H 8L6 Canada.
The Physicians Services Incorporated Foundation (Grant # 95-33) and the Canadian Orthopaedic Foundation (Hip Hip Hooray) provided grant support in assistance of this project.
Many authors have evaluated the accuracy of plain film radiographs for assessing articular cartilage wear in osteoarthritis of the knee, and efforts at enhancing the sensitivity of X-rays have been reported. Though magnetic resonance imaging has been shown to be the most sensitive tool for assessing articular cartilage, 7,17 plain film radiography remains the most widely available and least costly imaging method. Optimizing the information from radiographs could decrease the need for alternative imaging techniques and promote more efficient use of these resources.
The flexion view for knee radiography was first described by Holmblad, 20 and Resnick et al. 29 showed that a standard tunnel view was more sensitive to joint space narrowing than standing anteroposterior (AP) projections. Maquet has reported biomechanical data that demonstrated that peak articular stresses were found on the femorotibial articulation at 28° flexion because of diminished contact area. 24
Rosenberg 30 reported on the 45° posteroanterior (PA) weight-bearing projection in a consecutive series of 53 patients who underwent arthroscopy. They found that severe articular cartilage changes of ulceration or erosion were more consistently detected on the 45° projection (80–85%) than on standing AP radiographs (25–30%). Most have not differentiated the radiographic accuracy as a function of compartment localization.
There were several goals of the current study. One, to confirm that flexion weight bearing was more accurate than erect standing in predicting full thickness articular cartilage wear. Second, to determine if there were compartmental differences (medial versus lateral) in the utility of these radiographs. We also sought to evaluate whether mechanical alignment was more sensitive than joint space narrowing as an indicator of full thickness cartilage loss.
All patients ages 40 to 75 referred to the orthopedic outpatient clinic between March 1995 and November 1997 with osteoarthritis of the knee (as defined by the American Rheumatological Association 2) were considered for eligibility for the current study. Patients with inflammatory or traumatic forms of osteoarthritis were excluded. The age criteria were chosen in accordance with most therapeutic studies of osteoarthritis. Patients who remained symptomatic despite supervised physical therapy and comprehensive medical management were considered for arthroscopy. The failed conservative management included oral and topical analgesics, nonsteroidal antiinflammatory medications, and intraarticular cortisone injection in some patients, and was generally supervised by the referring physician. Most were guided by their impression of mechanical symptoms and swelling as best indications for surgery and possible benefit. The study was explained to all prospective patients and their informed consent was required for participation in the study, which was reviewed and approved by the Research Ethics Board of the Ottawa General Hospital.
Radiographs were obtained 1 week preoperatively in both the 3-foot standing AP and a 45° flexion weight-bearing PA projection. 30 The 45° PA flexion radiograph was made with the femur angled 25° and the tibia 20°, while the X-ray was centered at the superior pole of the patella and directed 10° caudad (Figure 1). The latter has been suggested as a more sensitive technique for joint space narrowing. A foot map was used to normalize rotation for both views, and a foam wedge was used to control flexion for the 45° X-ray projection. Fluoroscopy was not used to aid in positioning. This standardized technique was felt to be essential to enhance the reproducibility of the measurement. Joint space and knee alignments were recorded. Joint space in millimeters was measured for each tibiofemoral compartment at the narrowest point of contact. Tibiofemoral alignment was measured as described by Moreland, 26 using a standard goniometer. All radiographs were assessed by two orthopedic surgeons (G.F.D. and R.J.F.), who were blinded to the arthroscopic and clinical manifestations of the patients. We had agreed on the format for evaluating the radiographs before proceeding with the study. Interobserver agreement was described by both the overall percent agreement and the kappa coefficient. 22
All subjects underwent arthroscopy of the knee under general or spinal anesthesia. All surgery in this study was performed via standard anterolateral and anteromedial skin portals. The use of tourniquet was optional. A complete tricompartmental diagnostic arthroscopy was first carried out and videotaped prior to any intervention to permit grading of articular cartilage surfaces (see below). The surgeon then performed the procedure, which included resection of loose chondral flaps, unstable meniscal tears, and synovectomy only where required for visualization.
Arthroscopic Classification of Articular Cartilage Damage
Dougados et al. 16 devised a classification scheme for each tibiofemoral and patellofemoral compartment, which included depth of lesion, surface area involvement, and exact location. The description of the depth of lesions was as follows: Grade I, softening and swelling of cartilage; Grade II, superficial fissuring of the cartilage surface, velvet-like appearance; Grade III, deep fissuring reaching subchondral bone, including partially detached chondral flaps or crab-meat-like appearance; and Grade IV, erosion to exposed bone. Surface area involvement was estimated as a percentage of compartment involvement, and location of the lesion was recorded on an articular diagram.
This classification was subsequently validated in a population of patients fulfilling the American College of Rheumatology clinical and radiographic criteria for osteoarthritis of the knee, 2,3 similar to the patients enrolled in the current study. We therefore selected the French Society of Arthroscopy (SFA) scoring scheme for articular scoring, and chose to use “grades” rather than “scores,” as Ayral et al. 3 showed there was complete agreement in 9 of 10 cases between the same or different observers when using grades as a score.
Interobserver agreement of all clinical variables was measured with a kappa coefficient (κ). 22 Statistical analysis was performed using SPSS for Windows v. 6.1.3 (SPSS, Inc., Chicago, IL, U.S.A.).
Cross-tabulation by chi-squared test was used to determine the association of articular cartilage grade against joint space measure and mechanical alignment, respectively, for each compartment. Joint space measure was then dichotomized into several cut-points between 0–7 mm for each compartment to determine the classification performance for the prediction of severe (Grade IV) articular cartilage wear. The cut-point that provided the greatest accuracy (highest percentage of true positives and true negatives) was chosen for the subsequent comparison of the two radiographic views.
Two hundred seven patients were referred for admission into the study once the participating surgeon had scheduled arthroscopic surgery. Forty did not meet the criteria for established osteoarthritis as defined, and 15 were excluded because of unsatisfactory video recording, which could not be reliably interpreted postoperatively by the assessor blinded to the patient's preoperative criteria. One hundred fifty-two patients with mean age 60.5 years (± 8.5 SD), of whom 51% were female, comprised the study cohort. Pain was the most common presenting complaint, particularly with stair climbing and arising from a chair. The cohort demonstrated a wide range of self-reported disability as measured by the baseline Western Ontario and MacMaster (WOMAC) pain score of 24.2 ± 10.5 (mean ± SD) with a range of 3–49 on a visual analog scale that spanned 0–50 mm. 4 Ninety-two patients had an unstable meniscal tear as determined independently by the principal investigator following videotape review. The vast majority were degenerative tears.
Figure 2 localizes the distribution of chondral damage severity graded arthroscopically for all cases. The medial compartment was considerably more damaged, with 57% showing Grade III or IV involvement. In contrast, both the lateral and the patellofemoral compartments were less severely involved with only 13% and 17% Grade III and IV changes, respectively.
Accordingly, the spectrum of articular wear severity was more evenly distributed for the medial compartment than for the other two compartments. The distribution of pain symptoms and arthroscopic severity suggests that the cohort captured the spectrum of disability and disease severity in the medial compartment.
The coronal tibiofemoral 26 measures for the 3-foot standing AP film showed a trend toward knees with varus alignment. The 25th, 50th, and 75th percentile values were 2, 4, and 5°, respectively. Varus alignment was defined as 3° or less, neutral as between 4–8°, and valgus as 9° or more, based on values established by Moreland with healthy adult males. 26
The kappa coefficient is calculated as the percentage agreement expected beyond that of chance and is hence a better index of the reliability of the variable being studied. 31 The kappa coefficient is a more valid measure of concordance because, with the exception of extremes of prevalence, the kappa coefficient is not as sensitive to the underlying prevalence of the variable being measured as would be the percent agreement. Adopting the standard adopted by Landis, 22 the interpretation of kappa coefficients is commonly described as follows: (0–0.2 = slight, 0.2–0.4 = fair, 0.4–0.6 = moderate, 0.6–0.8 = substantial, and 0.8–1.0 = almost perfect). Reliability was moderate to substantial for joint space narrowing using 2 mm as a cutoff for either tibiofemoral compartment in each radiographic projection (Table 1). Agreement was only fair for mechanical limb alignment using the techniques and descriptions defined by Moreland. 26
Prediction of Severe Articular Chondropathy
Forty-six patients were classified with severe Grade IV medial compartment chondropathy at arthroscopy. Thirty-two patients had full thickness defects on both surfaces, 8 patients on the tibia, and 6 on the femur only, respectively. There was little difference in the average joint space height measured by the 45° PA view (1.4 ± 1.4 mm SD) or the 3-foot standing AP view (1.9 ± 1.6 mm SD). In both instances, a cut-point of 2 mm appears to have the best combination of true and false positive rates with the corresponding classification characteristics given in Table 2. On the basis of this evaluation, no significant advantage could be found for either view in evaluating the medial compartment severity (Fig. 3). Using a 2-mm side-to-side difference as reported by Rosenberg, 30 the sensitivity dropped from 78% to 47%, so that in this cohort the absolute joint space measure was more pertinent to severe chondropathy than comparison to the contralateral side.
Twelve patients were categorized as having severe lateral compartment articular chondropathy (Grade IV) at the time of arthroscopy. The lateral joint space height averaged 1 .0 ± 1.7 mm on the 45° PA radiograph, compared with 2.7 ± 1.1 mm on the 3-foot standing AP view. Figure 4 shows a typical case.
A review of all possible cut-points showed that 2 mm or less was most useful with the 45° PA view being much more sensitive (83% versus 42%) at correctly detecting the most severe chondropathy (Table 3). Ten of 12 cases with severe chondropathy were correctly identified by the 45° PA view. The two cases misclassified as false negatives had joint space heights of 3 and 5 mm, and differed in that the articular wear was confined to the medial side of the lateral tibial plateau without changes on the femoral side. As such, these were not typical cases of lateral compartment osteoarthritis, but received Grade IV grading according to the SFA scale. Two other cases (both false positive) were incorrectly classified in the severe chondropathy group. Despite these exceptions and small number of cases, the 45° PA weight-bearing view was of greater value for predicting severe articular wear in the lateral compartment, largely because of increased sensitivity.
Using a similar methodology as that for joint space narrowing to define optimal cut-points for predicting Grade IV chondral changes, mechanical limb alignment was less accurate than the latter for either compartment. In addition, the relatively poor interobserver agreement (κ = 0.29) makes this view even less reliable for predicting Grade IV changes except in the most severe cases, as in tibiofemoral alignment <2° (for the medial compartment).
A number of views have been advocated for improved diagnostic accuracy of osteoarthritic knees, although it remains difficult to consistently reproduce positioning to get images that can be compared for comparative or longitudinal follow-up. 10 Without precise control of the X-ray beam, the latter cannot consistently lie parallel to the articular surfaces and perpendicular to the film. The variability in tibial slope between 0 and 20° further complicates efforts at standardizing plain film radiographic imaging. Ravaud et al. 28 showed that changes of X-ray beam inclination as small as 5–10° created joint space width differences of 0.43–1.53 mm. Others have shown spurious results due to minor degrees of flexion. Fife et al. reported that increased flexion alone can result in apparent joint space loss of up to 25% in the medial compartment. 18 In contrast, Ravaud et al. found that joint space width was increased up to 12.5% with only 10° of flexion. They also emphasized the importance of standardizing foot rotation, beam inclination, and direction. Other efforts to more precisely estimate articular cartilage wear involve the use of fluoroscopy for reproducible joint positioning and digitizing radiographic images for computerized assessment of joint space width and osteophyte presence. 13–15 Recognizing the limitations of standard radiography, Buckland-Wright has reported on the use of microfocal radiography for the evaluation of changes in osteoarthritis of the knee. 8–11 Microfocal X-ray units have a much smaller X-ray source (<15 μm) than standard units (0.3–1 mm), which allows higher magnification and greater detail. The object is placed close to the source (20–30 cm) and the image projected at a distance of 1.5–2 m, with resulting magnifications by a factor of 4–10. The increased distance reduces secondary scatter, increasing the contrast and detail such that objects of 25–50 μm can be detected. Combining this imaging with precise and reproducible limb positioning increases the precision and accuracy of joint space determination in osteoarthritic knees. 12 Adoption of this would appear to be potentially more accurate than the standard 45° PA view described in this study.
The clinical reality at this time for orthopedic surgeons is that only Outerbridge lesions of Grades III–IV 27 are addressed with contemporary surgical techniques. Thus, despite the limitations of standard radiography, they remain for most clinicians the most practical, readily available and useful imaging tools for assisting with clinical management. Not surprisingly, the availability of arthroscopy has permitted evaluation of plain film radiography for detecting articular cartilage wear. Lysholm et al. 23 compared AP weight-bearing views graded by Ahlback's scale 1 with arthroscopic examination in patients with varus osteoarthritis. While X-rays were sensitive to medial cartilage loss, they were unreliable in detecting lateral compartment chondral damage. Eight of nine patients with lateral compartment chondral disease had normal radiographic scores for that compartment. Fife et al. 18 also showed disappointing results with standing AP radiographs. Theirs was a younger cohort with a mean age of 36 years, and only 15 of 161 patients had Grade IV chondropathy. Rosenberg 30 described a 45° posteroanterior weight-bearing projection in a consecutive series of 53 patients who underwent arthroscopy. Using a minimum of 2 mm difference in joint space width as a criterion, they demonstrated greater sensitivity for articular cartilage ulceration or erosion with the 45° PA projection (80–85%) than with standing AP radiographs (25–30%). There were no false positives for either the medial or lateral compartments in their study. Although this classification rate is superior to that of the current study, it requires that the contralateral compartment presumably be uninvolved since it relies on comparative narrowing of the joint space. They did not provide data as to the interobserver agreement for the measurements. The 2 mm side-to-side difference was much less sensitive in the current study, as several patients had bilateral disease rendering this measure less helpful. Messieh et al. 25 showed narrowing of 2 mm or more in 32 of 198 tibiofemoral compartments, but did not provide data for the accuracy of either measurement with their standing tunnel view at 30° flexion when compared with AP standing X-ray.
There is still debate on how best to define radiographic osteoarthritis for research and clinical purposes. Should the condition be defined on presence of osteophytes, joint space narrowing, or both? A commonly used scale was first proposed by Kellgren and Lawrence to grade severity of knee osteoarthritis. 21 Criticized because it was derived from nonweight-bearing projections, the classification emphasizes the presence of osteophytes, which some argue are often a natural occurrence with aging and not always pathologic. 19 Joint space narrowing is only a gross descriptor in this scheme. Blackburn et al. 5 attempted to correlate radiographic changes with arthroscopic findings using the Kellgren-Lawrence scale, which, not surprisingly, underestimated the articular cartilage damage because of limitations in the scale and of the standing AP projection. Brandt et al. 6 recently offered a new classification based on AP standing projections that included joint space narrowing as a criterion. They described joint space narrowing as a percentage loss, presumably compared with the contralateral normal similar compartment. Despite this attempt at using joint space narrowing as a discriminator, they reported their scale significantly underestimated the degree of cartilage loss in 27% of patients. Admittedly this was a selected population (80% male) of younger patients (mean age 40) with relatively mild osteoarthritis as seen in a sports medicine clinic and is not likely representative of the nontraumatic degenerative older arthritic patient. They emphasized that neither osteophytes nor joint space narrowing were accurate indicators for early osteoarthritis.
The 45° PA view proved to be as sensitive as the 3-foot standing anteroposterior view for predicting full thickness articular cartilage loss in the medial compartment and more sensitive in the lateral compartment. There are other issues that we believe favor routine adoption of the 45° PA view as a screening radiograph for osteoarthritis. The technical and professional charge of the 45° PA view is $(CAN)20.25 versus $(CAN)30.23 for the 3-foot standing view according to the Ontario Ministry of Health's 1998 Schedule of Benefits. A regular sized view-box is adequate for the 45° PA view allowing easier use. The presence of the other knee on the same projection allows for ease of comparison for the clinician and education for the patient to graphically depict asymmetries of joint space or the presence of osteophytes. This view can be duplicated in smaller radiology departments of community hospitals and outpatient clinics where the majority of these patients are initially evaluated. One disadvantage is the inability to calculate the mechanical weight-bearing axis as with the 3-foot standing view. Nevertheless, it can be argued that the latter can be ordered on an individual basis and is more pertinent when more advanced procedures such as osteotomy and knee arthroplasty are being considered, usually in larger centers with full orthopedic coverage and larger radiology departments. Finally, other techniques such as microfocal radiography or magnetic resonance imaging may be more appropriate for monitoring longitudinal progression because of greater sensitivity to small changes in joint space width.
We recommend the 45° PA view as a screening radiograph for osteoarthritis of the knee in a primary care setting. In this population, joint space height of 2 mm or less predicted severe Grade IV articular wear in the medial and lateral compartment with accuracies of 77% and 95%, respectively.
The authors thank the following orthopedic surgeons for their participation in the study and the contribution of their patients: J. Brunet, MD, J. Bouchard, MD, J. P. Desjardins, MD, R. Feibel, MD, A. Giachino, MD, J. McAuley, MD, and G. Moreau, MD. Thanks also to Kim Morin who helped with preparation of the manuscript.
1. Ahlback S. Osteoarthrosis of the Knee. A Radiographic Investigation
. Stockholm: Karolinska Institutet, 1968:11–15.
2. Altman R, Asch E, Bloch Dea. Development of criteria for the classification and reporting of osteoarthritis. Arth Rheum
3. Ayral X, Dougados M, Listrat V, et al. Arthroscopic evaluation of chondropathy in osteoarthritis of the knee. J Rheum 1996; 23:698–706.
4. Bellamy N, Buchanan WW, Goldsmith CH, et al. Validation study of WOMAC: a health status instrument for measuring clinically-important patient—relevant outcomes following total hip or knee arthroplasty in osteoarthritis. J Orthop Rheum 1988; 1:95–108.
5. Blackburn W, Bernreuter W, Rominger M, et al. Arthroscopic evaluation of knee articular cartilage: A comparison with plain radiographs and magnetic resonance imaging. J Rheum 1994; 21:675–679.
6. Brandt K, Fife R, Braunstein E, et al. Radiographic grading of the severity of knee osteoarthritis: relation of the Kellgren Lawrence grade to a grade based on joint space narrowing, and correlation with arthroscopic evidence of articular cartilage degeneration. Arth Rheum 1991; 34:1381–1386.
7. Broderick L, Turner D, Renfrew D, et al. Severity of articular cartilage abnormality in patients with osteoarthritis: evaluation with fast spin-echo MR versus arthroscopy. AJR 1994; 162:99–103.
8. Buckland-Wright JC. A new high-definition microfocal X-ray unit. Br J Radiology 1989; 62:201–288.
9. Buckland-Wright JC. Quantitative radiography in osteoarthritis: microfocal radiography. Baillieres Clin Rheum 1996; 10:415–420.
10. Buckland-Wright JC. Quantitative radiography of osteoarthritis. Ann Rheum Dis 1994; 53:268–275.
11. Buckland-Wright JC, Macfarlane DG, Jasani MK, et al. Quantitative microfocal radiographic assessment of osteoarthritis of the knee from weight bearing tunnel and semiflexed standing views. J Rheum 1994; 21:1734–1741.
12. Buckland-Wright JC, Macfarlane DG, Williams SA, et al. Accuracy and precision of joint space width measurements in standard and macroradiographs of osteoarthritic knees. Ann Rheum Dis 1995; 54:872–880.
13. Conrozier T, Vignon E. Quantitative radiography in osteoarthritis: computerized measurement of radiographic knee and hip joint space. Baillieres Clin Rheum 1996; 10:429–433.
14. Dacre JE, Coppock JS, Herbert KE, et al. Development of a new radiographic scoring system using digital image analysis. Ann Rheum Dis 1989; 48:194–200.
15. Dacre JE, Huskisson EC. The automatic assessment of knee radiographs in osteoarthritis using digital image analysis. Br J Rheum 1989; 28:506–510.
16. Dougados M, Ayral X, Listrat Veal. The SFA system for assessing articular cartilage lesions at arthroscopy of the knee. Arthroscopy
17. Drape JL, Pessis E, Auleley GR, et al. Quantitative MR imaging evaluation of chondropathy in osteoarthritic knees. Radiology 1998; 208:49–55.
18. Fife RS, Brandt KD, Braunstein EM, et al. Relationship between arthroscopic evidence of cartilage damage and radiographic evidence of joint space narrowing in early osteoarthritis of the knee. Arthr Rheum 1991; 34:377–382.
19. Hernborg J, Nilsson B. The relationship between osteophytes in the knee joint, osteoarthrosis and aging. Acta Orthop Scand 1973; 44:69–74.
20. Holmblad E. Postero-anterior X-ray view of the knee in flexion. JAMA 1937; 109:1196–1197.
21. Kellgren J Lawrence J. Radiological assessment of osteoarthrosis. Ann.Rheum Dis 1957; 16:494–501.
22. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159–174.
23. Lysholm J, Hamberg P, Gillquist J. The correlation between osteoarthrosis as seen on radiographs and on arthroscopy. Arthroscopy 1987; 3:161–165.
24. Maquet P, Van de Berg A, Simonet J. Femorotibial weight-bearing areas. Experimental Determination. J Bone Joint Surg (Am)
25. Messieh S, Fowler P, Munro T. Anteroposterior radiographs of the osteoarthritic knee. J Bone Joint Surg (Br)
26. Moreland JR, Bassett LW, Hanker GJ. Radiographic analysis of the axial alignment of the lower extremity. J Bone Joint Surg (Am)
27. Outerbridge R. The etiology of chondromalacia patellae. J Bone Joint Surg (Br)
28. Ravaud P, Auleley G-R, Chastang C, et al. Knee joint space width measurement: an experimental study of the influence of radiographic procedure and joint positioning. Br.J Rheumatol. 1996; 35:761–766.
29. Resnick D, Vint V. The “tunnel” view in assessment of cartilage loss in osteoarthritis of the knee. Radiology 1980; 137:547–548.
30. Rosenberg T, Paulos L, Parker R, et al. The forty-five-degree posteroanterior flexion weight-bearing radiograph of the knee. J Bone Joint Surg (Am)
31. Sackett DL, Haynes RB, Guyatt GH, et al. Clinical Epidemiology. A Basic Science for Clinical Medicine
. 2nd ed. Boston: Little, Brown and Co., 1991;25–31.
Keywords:© 2001 Lippincott Williams & Wilkins, Inc.
Radiograph, standing; Radiograph, anteroposterior; Osteoarthritis; knee